EEE
TEE rrr
Table 1 Edge matching results (values in [um given in object space)
number RMS minimum | maximum SY RMS minimum | maximum SZ
of frames DXY DXY DXY DXZ DXZ DXZ
4 5.8 -12.0 14.8 15.4 13.1 -40.1 +27.4 284
6 5:9 -15.0 +11.4 13.6 13.5 -35.0 +25.3 233
8 6.6 -17.7 +14.9 14.3 12:5 -36.1 439.5 27.8
matching fails, because the adjustment solution does 6. ACKNOWLEDGEMENTS
not converge.
e The use of more than 4 images does not result in a
significant improvement of the matching results if the
camera configuration is good.
e The number of iterations per matched point using the
edge tracking procedure depends on the curvature of
the edge and the tracking step width. For the examples
with 4 images the average number of iterations was 3.
e The accuracy of LSM also depends on the design of
the template. If a very steep edge template (grey level
ramp, one pixel wide) is used, the radiometric
corrections will also cause the patch patterns to be
compressed into very steep gradients. This leads to
inaccurate LSM results.
The use of a larger patch size (7 x 7 or 9 x 9 pixel)
smooths the difference curve, because for the edge
matching more grey values are used along the edge.
However, the improvement in the results is not
significant.
The computing performance is 6.7 points/sec, measured
on a Sun 4/490 workstation and relates to an average
number of 4.3 iterations per matched point. The
performance was measurcd with the program version
without screen display using 4 images. 652 object points
were thus matched in a total time of 97.3 sec.
5. CONCLUSIONS
As evidenced by the controlled practical test presented in
this paper our algorithm has the capability to measure
natural object edges with a relative accuracy of 1:25000.
This translates into 0.023 pixel in image space and was
obtained with standard image acquisition hardware and
under non-optimized camera orientation conditions. A
further improvement can be expected through the use of
higher resolution CCD-imager chips, better signal
transfer and digitisation hardware and more accurate
camera orientations.
However, in order to obtain such high accuracies a
number of prerequisites must be met, like the existence of
a sufficiently well-defined object edge, an appropriate
illumination, and the usage of more than two CCD
frames. At this level, a close-range vision system can
well compete with a CMM (Coordinate Measurement
Machine), having the additional advantages of higher
processing speed and non-contact measurement
characteristics.
The software for this vision system was partially
developed under a contract with Standard Aero Limited,
and in cooperation with NCR, Canadian Institute of
Industrial Technology, both Winnipeg, Canada. We are
grateful for the support and cooperation.
7. REFERENCES
Baltsavias, E. P., 1992: Multiphoto Geometrically Con-
strained Matching. Ph. D. Dissertation, Mitteilungen Nr.
49, Institute of Geodesy and Photogrammetry, ETH Zu-
rich.
Gruen, A., 1985: Adaptive least squares correlation - a
powerful image matching technique. South African Jour-
nal of Photogrammetry, Remote Sensing and Cartogra-
phy, 14 (3), pp. 175-187.
Gruen, A.W., Baltsavias, E.P., 1988: Geometrically Con-
strained Multiphoto Matching. Photogrammetric Engi-
neering and Remote Sensing, Vol. 54, No. 5, May 1988,
pp. 633-641.
Gruen, A., Stallmann, D., 1991: High accuracy edge
matching with an extension of the MPGC-matching algo-
rithm. SPIE, Vol. 1526, Industrial Vision Metrology,
Winnipeg, 1991, pp. 42-55.
Gruen, A., Beyer, H., 1991: DIPS II - Turning a Standard
Computer Workstation into a Digital Photogrammetric
Station. ZPF - Zeitschrift für Photogrammetrie und
Fernerkundung, 1, pp. 2-10.
El-Hakim, S.F., 1990: Some solutions to vision dimension-
al metrology problems. SPIE, Vol. 1395, Close-Range
Photogrammetry Meets Machine Vision, Zurich, pp. 480-
487.
Raynor, J.M., Seitz, P., 1990: The technology and practical
problems of pixel-synchronous CCD data acquisition for
optical metrology applications. SPIE, Vol. 1395, Close-
Range Photogrammetry Mects Machine Vision, Zurich,
pp. 96-103.
